ABSTRACT
Despite being one of the most common indications for surgery, data on the types of meniscus tear that should be treated surgically are limited. Improving patient selection requires agreement on meniscus tear description. This study evaluates a simple MRI tear classification system for inter-observer agreement. Knee MRI examinations from 57 subjects from the Osteoarthritis Initiative cohort were reviewed by two sub-specialty trained, musculoskeletal radiologists. Based on two pulse sequences, each meniscus was classified by: tear or no tear; location of tear in anterior, middle or posterior third or multiple thirds; and displaced or non-displaced radial, horizontal, longitudinal or complex tear pattern. A tear was defined as signal abnormality extending to the surface on at least two images and displacement as more than 2 mm of extrusion or separation measured orthogonal to the tear plane. Kappa, weighted Kappa and percentage agreement were calculated. For the medial meniscus, Kappa and percentage agreement estimates were, respectively: the presence of tear, 0.79 and 89.5%; tear with displacement, 0.70 (weighted Kappa) and 66.0%; tear description, 0.47 and 61.4%; tear location, 0.64 and 79.0%. For the lateral meniscus, estimates were: the presence of tear, 0.75 and 89.5%; tear with displacement, 0.81 (weighted Kappa) and 86.0%; tear description, 0.56 and 78.9%; tear location, 0.74 and 87.7%. The strength of agreement between readers was moderate to substantial underscoring the challenge of meniscus tear classification.